On 4/27/2016 1:41 PM, Christophe Gisquet wrote: > I guess this should be used sparingly, and with benchmark reports to > back the few cases.
The problem with measuring perf for likely/unlikely is that half the time it's either negligible or impossible to measure consistently. That's part of what I dislike it (besides being scarred for like from reading GNU source code). - Derek _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
